Back to Workshop's Home Page
| Project team: | Norman Paton1, Richard Cooper2, Jessie Kennedy3, Carole Goble1,Phil Gray2, Peter Barclay3, Adrian West1, Michael Smyth3, Tony Griffiths1 (RA), Jo McKirdy2 (RA), Andrew Dinn3 (RA) |
1 University of Manchester, 2 University of Glasgow, 3 Napier University |
Teallach is an EPSRC funded research project which has been running since October 1996. It is a collaboration between the universities of Manchester, Glasgow, and Napier, with industrial collaborations including IBM, ICL, British Geological Survey, and Criterion Software. The focus of our investigation lies in the systematic and generic support for the development of user interfaces to object-oriented database applications.
The principle aim of our research is the development of models, a software architecture, and tools which will facilitate the rapid development of user interfaces to ODMG compliant databases.
Our approach is based on the specification of a set of integrated data models in which to express the information from which user interfaces can be derived. These data models represent the following aspects of a database user interface: the tasks to be performed with the application; the application domain; the end-users of the application; the dialogue (an abstract description of user-application communication); and the presentational aspects of the user interface itself.
An important subsidiary project aim is to enable user interfaces to be generated automatically based on:
One consequence should be that we can provide multiple consistent user interfaces to the same application.
Such a computer generated system is likely to be only a "first-cut" and will require subsequent modification by a developer; our system will at least allow and at best assist with such modification.
So far we have developed a first version of our data models and tested them in a simple case study. We are now engaged in a modification and refinement of our models. Among the issues we are addressing during this phase of the project and which will be the discussed at the workshop are:
Although we have already specified a dialogue meta-model, we believe the information it contains might be better located in either the task and/or presentation models, simplifying our overall modelling framework. We are currently investigating the advantages and disadvantages such a simplification.
We would like our system to be able to handle ad hoc (i.e. user-generated) OQL queries and the subsequent presentation of query results. This problem poses questions at both domain and presentation levels, affecting the required modelling components and the way the two models interact.
The set of models which describe a particular system are subject to change (reconfiguration), as is the user interface with which they are associated. A major challenge is to ensure that changes are communicated among models and to/from the user interfaces, as appropriate, and to ensure that consistency is maintained amongst these components.
The problem of handling communication and consistency is exacerbated by the fact that we impose no restrictions on the sequencing of model specification. Previous model-based approaches have only allowed for one-way generation and propagation of updates (e.g. from the domain model to the presentation model). We propose to relax this constraint.
How can our models be sufficiently "constructionally-oriented" that they can be used to generate the run-time system and sufficiently "design-oriented" that they can be built and modified by (human) system developers? This requires examining the levels of abstraction of the models and the information structures they contain.